neural network modelling

In this paper, the large inelastic deformation and failure mechanism of single and multi-layered circular plates under repeated uniform impulsive loading were studied. The ballistic pendulum was used to conduct a series of experiments (67 experiments) on aluminum alloy plates with different structural configurations. Three different layering configurations including single, double, and triple-layered plates made of the same material were considered and tested for the range of charge masses from 1.5g to 12.5g up to five times for repeated loading. The experimental results indicated large plastic global deformation with thinning happening at the clamped boundary and also tearing for some experiments. The results also represented that the maximum permanent deflections of plates were increased by the increase of the charge mass and the number of blast loads. On the other hand, the progressive deflection of the plates at the center was decreased exponentially with increasing the number of blasts. Furthermore, in the numerical modeling section, the Group Method of Data Handling (GMDH) neural network was used to present a mathematical model based on dimensionless numbers to predict the maximum permanent deflection of single and multi-layered circular plates under repeated impulsive loading. In order to increase the prediction capability of the proposed neural network for this process, the experimental data were divided into two training and prediction sets. Good agreement between the proposed model and the corresponding experimental results is obtained and all and 77% of data points are within the <10% error range for single and multi-layered plates, respectively.